Individuality in Piano Performance Depends on Skill Learning

Abstract

Expert musicians' performances embed a timing variability pattern that can be used to recognize individual performance. However, it is not clear if such a property of performance variability is a consequence of learning or an intrinsic characteristic of human performance. In addition, little evidence exists about the role of timing and motion in recognizing individual music performance. In this paper we investigate these questions in the context of piano playing. We conducted a study during which we asked non-musicians to perform a musical sequence at different speeds. Then we tested their learning performance at a fixed tempo. Focusing on the possibility to identify the participant based on performance features of timing and motion variability, we show that participant classification increases with practice. This suggests that 1) the individual timing signatures are affected by learning and 2) timing and motion variability is structured. Moreover, we show that motion features better classify individual performances than timing features.

Audio cues enhance mirroring of arm motion when visual cues are scarce

“…Although this is a possibility we could only rule out completely by testing the same subjects de novo under multiple conditions, the significant increase in typical performance over many subjects implies that our experimental conditions are changing behaviour. Thus, these dynamic signatures provide evidence that the highest performing individuals are better at learning to use information from the audio cues to mirror the avatar, enabling them to perform well in regions of parameter space [21,28,29]. The enhanced stability of mirroring runs that we observe is consistent with studies showing that short training sessions reduce error rates and temporal variability in motor tasks [16].…”

Computational Models of Expressive Music Performance: A Comprehensive and Critical Review

“…Here, for example, algorithms that can determine the difficulty of a piece (Sébastien et al, 2012;Nakamura et al, 2014b) or propose appropriate fingering strategies (Al Kasimi et al, 2007;Nakamura et al, 2014b;Balliauw et al, 2015) would be useful. Furthermore, computational models might help determine a performer's skill level (Grindlay and Helmbold, 2006;Caramiaux et al, 2017). Musical e-learning platforms such as Yousician 7 and Music Prodigy 8 (and many more, as this is a rapidly growing business segment for start-ups) might benefit from models of performance to provide a more engaging experience, as well as to develop better musicianship.…”

“…Marchini et al (2013Marchini et al ( , 2014) study expressive performance in string quartets using a combination of music-only related expressive parameters, as well as bow velocity, a dimension of movement directly related to performed dynamics. Caramiaux et al (2017) assess whether individuality can be trained, that is whether the differences in performance style are related to development in skill and can thus be learned. Their results suggest that motion features are better than musical timing features for discriminating performance styles.…”